feat(instance-chat): full multimodal attachment support via OpenClaw bridge

After verifying that the OpenClaw gateway's chat.send WebSocket RPC
accepts an 'attachments' array (confirmed from openclaw/openclaw source
and documentation), implement end-to-end image/file attachment support
for instance chat:

Bridge (openclaw-client.ts):
- chatSendAndWait() now accepts optional `attachments[]` parameter
- Passes attachments to chat.send RPC only when non-empty

Bridge (index.ts):
- /task-async accepts `attachments[]` from request body
- Forwards to chatSendAndWait unchanged

Backend (agent.controller.ts):
- executeInstanceTask() accepts IT0 attachment format
  { base64Data, mediaType, fileName? }
- Converts to OpenClaw format { name, mimeType, media: "data:..." }
- Saves attachments to conversation history via contextService
- Forwards to bridge via bridgeAttachments spread

Flutter (agent_instance_chat_remote_datasource.dart):
- createTask() now includes attachments in POST body when present

Flutter (chat_page.dart):
- Reverted Fix 5 (disabled button) — attachment button fully enabled
  in instance mode since the bridge now supports it

Attachment format (OpenClaw wire):
  { name: string, mimeType: string, media: "data:<mime>;base64,<data>" }

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

it0_app/lib/features/agent_instance_chat/data/datasources/agent_instance_chat_remote_datasource.dart

@@ -29,7 +29,7 @@ class AgentInstanceChatDatasource implements ChatRemoteDatasource {
       data: {
         'prompt': message,
         if (sessionId != 'new') 'sessionId': sessionId,
-        // Note: attachments are not yet supported for instance chat
+        if (attachments != null && attachments.isNotEmpty) 'attachments': attachments,
       },
     );
     return response.data as Map<String, dynamic>;

it0_app/lib/features/chat/presentation/pages/chat_page.dart

@@ -720,16 +720,10 @@ class _ChatPageState extends ConsumerState<ChatPage> {
               if (!isStreaming)
                 Padding(
                   padding: const EdgeInsets.only(left: 4),
-                  child: Tooltip(
+                  child: IconButton(
-                    message: widget.agentName != null
+                    icon: const Icon(Icons.add_circle_outline, size: 22),
-                        ? '附件功能暂不支持智能体对话'
+                    tooltip: AppLocalizations.of(context).chatAddImageTooltip,
-                        : AppLocalizations.of(context).chatAddImageTooltip,
+                    onPressed: isAwaitingApproval ? null : _showAttachmentOptions,
-                    child: IconButton(
-                      icon: const Icon(Icons.add_circle_outline, size: 22),
-                      onPressed: (isAwaitingApproval || widget.agentName != null)
-                          ? null
-                          : _showAttachmentOptions,
-                    ),
                   ),
                 ),
               Expanded(

packages/openclaw-bridge/src/index.ts

@@ -145,6 +145,12 @@ app.post('/task-async', async (req, res) => {
   const timeoutSeconds: number = req.body.timeoutSeconds ?? 120; // 2 min default for async tasks
   const idempotencyKey: string = req.body.idempotencyKey ?? crypto.randomUUID();
   const callbackData  = req.body.callbackData ?? {};
+  // Optional attachments — passed through to OpenClaw chat.send unchanged.
+  // Expected format: [{ name, mimeType, media }] where media is a data-URI.
+  const attachments: Array<{ name: string; mimeType: string; media: string }> | undefined =
+    Array.isArray(req.body.attachments) && req.body.attachments.length > 0
+      ? req.body.attachments
+      : undefined;
 
   // Return immediately — LLM runs in background
   res.json({ ok: true, pending: true });
@@ -162,6 +168,7 @@ app.post('/task-async', async (req, res) => {
     message: req.body.prompt,
     idempotencyKey,
     timeoutMs: timeoutSeconds * 1000,
+    attachments,
   }).then((reply: string) => {
     postCallback({ ok: true, result: reply, callbackData });
   }).catch((err: Error) => {

packages/openclaw-bridge/src/openclaw-client.ts

@@ -298,17 +298,23 @@ export class OpenClawClient {
     message: string;
     idempotencyKey: string;
     timeoutMs?: number;
+    /** Optional media attachments (images, PDFs, etc.) in OpenClaw format. */
+    attachments?: Array<{ name: string; mimeType: string; media: string }>;
   }): Promise<string> {
     const timeoutMs = params.timeoutMs ?? 30_000;
 
     // Send chat.send — resolves immediately with { runId, status: "started" }
+    const chatSendParams: Record<string, unknown> = {
+      sessionKey: params.sessionKey,
+      message: params.message,
+      idempotencyKey: params.idempotencyKey,
+    };
+    if (params.attachments && params.attachments.length > 0) {
+      chatSendParams['attachments'] = params.attachments;
+    }
     const ack = await this.rpc(
       'chat.send',
-      {
+      chatSendParams,
-        sessionKey: params.sessionKey,
-        message: params.message,
-        idempotencyKey: params.idempotencyKey,
-      },
       10_000, // 10s for the initial ack
     ) as { runId: string; status: string };
 

packages/services/agent-service/src/interfaces/rest/controllers/agent.controller.ts

@@ -827,7 +827,11 @@ export class AgentController {
     @TenantId() tenantId: string,
     @Req() req: any,
     @Param('instanceId') instanceId: string,
-    @Body() body: { prompt: string; sessionId?: string },
+    @Body() body: {
+      prompt: string;
+      sessionId?: string;
+      attachments?: Array<{ base64Data: string; mediaType: string; fileName?: string }>;
+    },
   ) {
     const instance = await this.instanceRepository.findById(instanceId);
     if (!instance) throw new NotFoundException(`Instance ${instanceId} not found`);
@@ -883,7 +887,7 @@ export class AgentController {
     await this.taskRepository.save(task);
 
     // Persist user message for display in conversation history
-    await this.contextService.saveUserMessage(session.id, body.prompt);
+    await this.contextService.saveUserMessage(session.id, body.prompt, body.attachments);
 
     // The OpenClaw bridge tracks conversation context internally via sessionKey.
     // We use our DB session ID as the key so each session has isolated context.
@@ -891,6 +895,15 @@ export class AgentController {
     const callbackUrl = `${process.env.AGENT_SERVICE_PUBLIC_URL}/api/v1/agent/instances/openclaw-app-callback`;
     const bridgeUrl = `http://${instance.serverHost}:${instance.hostPort}/task-async`;
 
+    // Convert IT0 attachment format → OpenClaw format for the bridge.
+    // IT0: { base64Data, mediaType, fileName? }
+    // OpenClaw: { name, mimeType, media: "data:<mimeType>;base64,<data>" }
+    const bridgeAttachments = body.attachments?.map((att) => ({
+      name: att.fileName ?? 'attachment',
+      mimeType: att.mediaType,
+      media: `data:${att.mediaType};base64,${att.base64Data}`,
+    }));
+
     this.logger.log(
       `[Task ${task.id}] Routing to OpenClaw instance ${instanceId} @ ${bridgeUrl}, session=${session.id}`,
     );
@@ -946,6 +959,7 @@ export class AgentController {
         idempotencyKey: task.id,
         callbackUrl,
         callbackData: { sessionId: session.id, taskId: task.id },
+        ...(bridgeAttachments && bridgeAttachments.length > 0 && { attachments: bridgeAttachments }),
       }),
     })
     .then(async (res) => {